CVE-2024-58400: ggml-org llama.cpp (High, CVSS 8.8)

What it is

llama.cpp before b3701 contains an out-of-bounds read vulnerability in llama_decode_internal and llama_encode_internal that allows remote unauthenticated attackers to read memory by supplying unvalidated token IDs. Attackers can send integer-array prompts to the llama-server /completion endpoint with IDs outside 0 to n_vocab to read embedding memory out of bounds or crash the process.

Description from the National Vulnerability Database.

What to do

Check whether you run ggml-org llama.cpp, then apply the vendor’s fix or mitigation from the advisories below.

Sources: NVD, FIRST EPSS, CISA KEV, and public exploit databases (availability only). ThreatRoundup reports whether public exploit code exists but never links to it: fake proof-of-concepts that infect security researchers are common. This page updates automatically.